This technique looks reminiscent of "superforecasting".
Caveat that their conclusion hinges on their assumptions, and we would expect that about 2/10 of their 80% predictions would actually be false if we could score them all.